[BE][MPS] Use convenience functions (#96521)
Introduce `getMPSScalarType(const Tensor&)` that calls `getMPSScalarType(t.scalar_type())`
And replace `getMPSScalarType(t.scalar_type)` with `getMPSScalarType(t)` throughout the codebase
Fixes #ISSUE_NUMBER
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96521
Approved by: https://github.com/seemethere
